Can local environmental constraints improve enterprise’s green innovation quality? Evidence from Chinese-listed firms

The target responsibility system of environmental protection is one of the vital channels to achieve a win–win situation for both economic development and environmental protection. Comprehensively investigating how local environmental target constraints drive enterprises to improve their green innovation quality is of enormous theoretical and practical significance for optimizing the implementation effect of environmental target constraints policies and boosting enterprises’ green and high-quality development. We empirically examine the mechanism of the impact of the intensity of different types of local environmental target constraints on the quality of corporate green innovation and the nonlinear relationship between them through innovatively constructing indicators of local environmental target constraint intensity and utilizing the knowledge width of green patents of listed companies as a proxy variable for enterprise green innovation quality. First, the strength of indirect environmental target constraints has a significant positive effect on the quality of green innovation, but further nonlinear characteristics reveal a significant inverse U-shaped relationship between them. Second, indirect environmental target constraint intensity has an inverted U-shaped trend in increasing the intensity of environmental regulation and influencing the digitalization of enterprises, which in turn forms an inverted U-shaped relationship with the quality of green innovation. Third, indirect environmental target constraint intensity works better in areas with policies prioritizing city over province, with mayors less than 57 years old, and for enterprises in technology-intensive industries.


Introduction
China made prominent progress in increasing its economic growth during the opening up over the past 40 years. However, the extensive economic development mode creates a heavy burden on the ecological environment. In April 2019, the State Development and Reform Commission and the Ministry of Science and Technology of China promulgated the Guiding Opinions on Establishing Market-oriented Green Technical Innovation System, wherein, the phrase "green technical innovation" appeared for the first time in a programmatic document by the China Communist Party and later appeared in the October 2021 Action Plan for Carbon Dioxide Emission Peaking Before 2030 issued by the State Council, which lists "innovations in green and low-carbon science and technology" as a major task. Thus, we understand that green innovation by enterprises is an important aspect of sustainable economic development and a critical path to realizing the carbon emission peak and carbon neutrality, as well as achieving high-quality economic development that provides both financial and ecological benefits. However, enterprises still face problems in green innovation quality because the novelty, inventiveness, influence, and practical value of their innovations are relatively low due to an imperfect patent system, distortion effects of some Responsible editor: Baojing Gu industrial incentive policies, and great difficulties and long cycles typical of innovation activities.
To accelerate sustainable economic development in China, the State Council promulgated the Decision on Implementing Scientific Development Outlook and Furthering Environmental Protection in December 2006, which stipulated that the KPIs of emission cuts would be taken into account when promoting or rewarding cadres. The Method for Assessing Reduction of Major Pollutants Emissions introduced the mechanisms of environmental protection accountability and the one-vote veto to assess local cadres' performance in reducing the emissions of major pollutants more strictly. Then, the local environmental target constraint system began to take shape and was further improved during the China's 12th and 13th 5-year periods. The local governments' environmental target constraints have Chinese characteristics and account for environmental protection when assessing local governments' political performance. These will directly influence the actions of local governments and impact enterprises' green innovations. To determine the effects of different local environmental target constraint intensities on enterprises' green innovation quality and whether the effects possess nonlinear features, we innovatively construct indicators of local environmental goal constraint intensity from the perspective of the environmental protection goal responsibility system, based on actual local government environmental goal constraint assessments. We build a theoretical model to study the influence of local environmental constraint intensity on enterprises' green innovation using the micro-data of green patent classification numbers from A-share-listed companies on the Shanghai and Shenzhen Stock Exchanges and the government work reports of cities for 2011-2019. We thus empirically examine the mechanism by which different local environmental target constraint intensities influence enterprises' green innovation quality at the micro-firm level, which is both theoretically and practically significant for realizing peak carbon emissions and carbon neutrality and achieving high-quality economic development.
This study offers several marginal contributions. First, in terms of the research approach and questions, our study based on the environmental accountability system, which has prominent Chinese characteristics, explores the influence of local environmental target constraint intensity on enterprises' green innovation quality and its nonlinear features. By clarifying the research questions and objects, this study not only complements the theoretical studies on the influence of Chinese environmental policies on enterprises' green innovations, but also sheds light on the differing influence of various types of environmental regulations on enterprises' green innovation quality, which is a further re-examination of the Porter hypothesis based on Chinese empirical evidence. Second, with respect to research objects, we follow the concept of "environmental target constraints" to create different indicators of local environmental target constraint intensities with reference to the environmental targets set in the government work reports of various cities and environmental performance released in public reports. We also use micro-data on listed companies' green patent classification numbers to measure their knowledge width and use it as a proxy variable for micro-enterprises' green innovation quality to augment prior studies on micro-enterprises' green innovation quality, which provides a reference for studying the quality of green innovation in micro firms. Third, after effectively identifying the influence of local environmental target constraint intensity on enterprises' green innovation quality, we adopt the macro-view of cities and the micro-view of enterprises to examine the mechanism of action between environmental regulations and science and technology (S&T) expenditure of the government and enterprises' media exposure, social responsibility, R&D investment, digitalization, and financing. Therefore, this study offers a new theoretical interpretation of how local environmental target constraints affect enterprises' green innovation quality.

Literature review
In terms of the impact of environmental policies on corporate green innovation, most studies are based on the Porter Hypothesis. Kammerer (2009) empirically concluded that environmental regulations not only drive innovation in green products and encourage extensive use of green products, but also promote their novelty, which lends solid support to the Porter Hypothesis (Cohen and Tubb 2018;Porter and Van Der Linde 1995). After the government implements suitable environmental supervision, enterprises may choose to adopt strategies to meet environmental supervision requirements and actively carry out green technical innovations to reduce pollutant emission, optimize production process, and effectively reduce pollutantdischarging costs (Rubashkina et al. 2015).
Most studies focused only on regional data indicators, such as environmental enforcement (Fan and Zhao 2019), reduced regional pollutant-discharge volume , and comprehensively weighted indicators of environmental governance investment (Wei et al. 2017). These studies explored the relationship between environmental regulations and green innovation at the meso-industrial and macro-regional levels. There are some scholars who have begun to adopt a microscopic perspective to study the influence of governments' environmental awareness on enterprises' green innovations in recent years. For example, Yang et al. (2021) and Zhang et al. (2022) took Chinese A-sharelisted companies from 2010 to 2019 and Chinese industrial firms as their research samples respectively to research the relationship between environmental regulations and enterprise green innovation. Fang et al. (2021) conducted research on the new Environmental Protection Law and found that it will put heavily polluting enterprises under supervisory pressure, potentially driving them to improve the quality of information disclosure and pursue green innovation. Liu et al. (2021) also found that enterprises filed more applications for more inventive and utility model environmental patents after the implementation of the new Environmental Protection Law. Sun et al. (2020) studied 132 enterprises from 16 heavily polluting industries and found that environmental regulations would encourage enterprises to invest more in environment-friendly R&D. However, some scholars oppose the Porter hypothesis, which states that environmental regulations inhibit corporate green innovation. Kemp and Pontoglio (2011) held that enterprises will switch focus to environmental management projects in order to meet governments' environmental requirements, which may incur higher costs and crowd out financial resources for R&D and technical innovations. He et al. (2021) concluded that environmental regulations will create a crowding-out effect on enterprises' R&D investment and negatively impact their patent output. Testa et al. (2011) also argued that environmental regulations such as pollutant-discharging fees will not benefit enterprises in carrying out technical innovations.
Therefore, an increasing number of scholars realize that the influence of environmental regulations on enterprises' green innovations might be nonlinear and heterogeneous (Rexhäuser and Ramer 2014). Chinese environmental regulations can be classified into three main categories, namely command-and-control, market-incentive, and voluntary (Chen and Monahan 2010;Andr´e et al. 2011;Xie et al. 2017). Wang et al. (2021) found that the influence of environmental regulations on enterprises' green innovations has an inverted U-shape. Jiang et al. (2018) defined environmental regulations from the two dimensions of industry and region and discovered the negative effects of industrial supervision on innovation performance, as opposed to regional regulation, which contributes to innovation performance. Li and Xiao (2020) comprehensively observed the incentive role of heterogeneous environmental regulations for encouraging enterprises to pursue green innovations and agreed that pollutant-discharging fees would force enterprises to conduct green innovation while environmental grants will have a crowding-out effect on enterprises' green innovation capability. Gao et al. (2021) also found that increasing pollution charges has a significant and positive effect on controlling industrial waste SO 2 . Tang et al. (2020) found that command-and-control environmental regulation negatively influences enterprise green innovation efficiency in a short term. Gao et al. (2022) found that market-based environmental regulations have a better improvement effect on green total factor energy efficiency than command-based environmental regulations. Li and Gao (2022) confirmed the incentive effect of market-based environmental regulations on firms' green innovation, while penalty had more significant influence on enterprises than subsidy at the early stage. Zhu et al. (2021) demonstrated that the direct effect of voluntary environmental regulation on technological innovation is significantly positive. Zhang et al. (2022) comprehensively summarized the differential impact of heterogeneous environmental regulations (command-and-control, marketincentive, and public-participation) on different types of corporate green innovation (cleaner production technology innovation and end-of-pipe technology innovation). Wu et al. (2021a, b) argued that the design and enforcement of environmental regulations must account for enterprises' actual conditions and differences to encourage them to pursue green innovations.
In summary, existing research has several gaps. First, the relationship between environmental policy and corporate green innovation has not yet reached a unified consensus, and there are significant differences in the effects of various types of environmental regulation on corporate green innovation. The Porter hypothesis is supported by foreign environmental regulatory policy systems dominated by market-based environmental regulations. In comparison, most of China's environmental regulations are currently mandatory (Wang et al. 2015a, b). Local officials not only make but also implement these mandatory environmental regulations. Currently, few studies account for both environmental policies and governments' environmental accountability assessments, and there is a lack of research examining the government's autonomous environmental goal constraints. Second, most of the current studies on definition and measurement of environmental regulation in China are broad and based on the macro-level, which does not directly reflect their Chinese characteristics. Studies using reduced pollutant-discharge volume and pollution treatment cost as proxy variables cannot evaluate and examine the effectiveness of different environmental regulations accurately, nor differentiate and summarize the heterogeneous influence of different environmental regulations. In particular, it is not suitable for matching with micro enterprises. Third, the scarce studies on micro-enterprises' green innovations always focused on the number of green patents. There are few studies that concentrate on enterprises' green innovation quality at the micro-level. The relationship between environmental policies and the quality of green innovation in firms has been poorly explored. Fourth, Yu et al. (2020) and Tao et al. (2021) empirically examined the influence of environmental accountability on enterprises' technical and green innovations; however, their studies focused only on macro-regulations and assessed their exogenous impacts based on whether the regulations were implemented. They did not subclassify or quantitatively measure the constraints of different environmental targets. There is much more scope to systematically and deeply understand the influence and incentive power of the intensities of various environmental target constraints on enterprises' green innovation quality. The indirect reaction mechanism is yet to be discovered, and the empirical design of its effect needs to be optimized urgently.

Theoretical model
We construct the theoretical model mainly by following the mechanism path by which local environmental target constraints will influence enterprises' green innovation quality by increasing the rigidity of environmental regulations (Yu et al. 2020). Environmental target constraints will motivate local governments to implement specific environmental regulatory policies. The main market-based environmental regulatory policies commonly adopted in China include the following: environmental taxes and emissions trading mechanisms. There are three scenarios in which environmental target constraints through environmental regulations affect enterprises' green innovation quality: (i) the influence of environmental taxes on enterprises' green innovation quality, (ii) the influence of auctioned emission rights on enterprises' green innovation quality, and (iii) the influence of emission rights trading on enterprises' green innovation quality. We borrow the theoretical framework of Xu et al. (2012), in which under the hypothesis of rational economic actors, enterprises maximize value and minimize costs under given incomes when facing local government environmental target constraints. Under this hypothesis, we explore the influence of local environmental target constraints on the quality of corporate green innovation through environmental taxes, emission rights for auctions, and emission rights trading, respectively. (Details of the theoretical model derivation process are shown in Appendix A.) According to the results of theoretical model derivation, when local government target constraint intensity increases, environmental tax rates and the prices of the auctioned emission rights increase under the optimal emission volume, which increases the expected net income from improved green innovation quality and stimulates enterprises to boost green innovation quality. When an increasing number of enterprises pursue green innovation quality, the price of tradable emission rights falls and the expected net income from improved green innovation quality decreases marginally and progressively, which weakens the incentive to pursue green innovation quality. Therefore, the intensity of local environmental target constraints will form an inverted U-shaped relationship with the quality of enterprise green innovation by influencing the intensity of environmental regulations.

Direct influence
When local governments at all levels make environmental targets public and assess their achievement, it reinforces the importance of environmental management in the political tasks of local governments, which in turn drives them to introduce environmental management measures, the most typical of which is the "closure and transfer" of enterprises . Facing stringent environmental controls and standards, enterprises with financial and technological capabilities will choose green innovations to transform and upgrade their development (Milani 2017). Innovation quality is key to enterprises' interests. On the one hand, to earn more from their green innovations, enterprises will cultivate unique competitive advantages and stimulate themselves. On the other hand, facing fierce market competition and strict government environmental supervision, enterprises under pressure from governmental environmental performance assessments and heterogeneous environmental regulations will honor their social responsibilities by enhancing the quality of their green innovations. Therefore, we propose: Hypothesis 1: Local governments' environmental target constraint intensity will significantly improve enterprises' green innovation quality.

Indirect influence
At the macro-mechanism level, Yu et al. (2020) held that local environmental target constraints will enhance the rigidity of environmental regulations to encourage technical innovation. However, the positive action of environmental target constraint intensity on enterprises' green innovation quality may change dynamically. First, although the implementation of command-based environmental regulations will have a certain promotion effect on enterprises' green innovation initially, due to the existence of penalty caps, it is likely that some enterprises will be satisfied with "meeting the standards" of green innovation activities, that is, the compensation benefits resulting from innovation activities will converge with the mandatory penalty costs caused by environmental noncompliance. Thus, there will be no effective incentive for enterprises to further improve the quality of green innovation. Command-based environmental regulations raised the environmental costs of the production process, which encroach on resources for green innovations (Petroni et al. 2019). Second, when incentive-based environmental regulations such as environmental taxes and grants improve, enterprises will increase their investment in green innovation (Bi and Yu 2016), which has a certain "resource compensation effect" (Li and Xiao 2020) on corporate green innovation, which in turn is conducive to improving the quality of green innovation. However, given the government's information disadvantage on enterprises' green innovation activities, responsive and strategic innovation behaviors arise. As the distortionary effect of resource allocation due to environmental taxes becomes greater than the ameliorating effect of negative environmental externalities, enterprises will reduce their R&D investment in green innovation when they reach a certain level (Zhang et al. 2017) or even engage in low-effort pursuits of a short-term increase in the quantity of green innovation and low-quality innovation to gain short-term green innovation quantity (Tao et al. 2021). In particular, when the marginal benefits generated by green innovation gradually decline (Xu et al. 2012), and the R&D investment and marginal cost for enterprises to maintain high-quality green innovation gradually increase, coupled with the increasing difficulty and risk of high-quality green innovation, the incentive for enterprises to further improve the quality of green innovation will decrease significantly.
Therefore, we propose: Hypothesis 2: By increasing the rigidity of environmental regulations, local environmental target constraint intensity creates an inverted U-shaped relationship with enterprises' green innovation quality.
Local officials, faced with the pressure of short-term performance appraisal, are more inclined to choose methodological initiatives that yield faster results in the short term to gain significant short-term results . However, it usually takes 3-5 years for a green innovation to develop from design to application, while it has a long cycle before improving environmental pollution and certain risks (Bian and Bai 2017), which is especially true for high-quality green innovations. Therefore, pressed by environmental target constraint assessments, local officials with elastic tenures tend to increase environmental protection expenditures that are more effective and shorter in the direct treatment of environmental pollution (Li and Bai 2021) and cut expenses for technical innovations that show unclear effects in tackling environmental pollution and are slow to translate into actual application. Therefore, increased local environmental target constraint intensity will crowd out the government's S&T expenses, and enterprises will face heightened costs to pursue high-quality green innovations, which may deter enterprises from enhancing the quality of green innovations. Therefore, we propose: Hypothesis 3: Increased local environmental target constraint intensity will cut the government's S&T expenses and inhibit enterprises from pursuing high-quality green innovations.
At the micro-mechanism level, the environmental accountability system guides the media and public to focus more on environmental protection (Tao et al. 2021), which puts enterprises under stricter environmental supervision and alerts enterprises' executives to related risks (Snyder and Stromberg 2010). Social opinions and media supervision force enterprises to invest more resources into green innovations (Fan and Fu 2021). Environmental initiatives and attention from the public and media will stimulate enterprises to increase investment in environmental protection (Wang et al. 2017), which will help enhance their sense of social responsibility and normalize their environmental behaviors (Han et al. 2021). By voluntarily fulfilling their social responsibilities to meet stakeholders' expectations for environmental protection, enterprises not only closely follow social values and regulations (Han et al. 2021), but also build their brands and images and elevate their performance (Wang et al. 2015a, b). Enterprises with enhanced awareness of social responsibility will invest more in R&D with prominent social benefits (Engida et al. 2020). Therefore, to meet both economic interests and social values, enterprises will set environmental targets and develop long-term green development strategies to voluntarily pursue breakthrough green innovations with higher quality. The institutional background of environmental target constraints and media opinion supervision, coupled with the national policy call for "promoting the deep integration of science and technology innovation with the real economy," creates a favorable economic and political environment and orientation to motivate enterprises to further increase their R&D investment to improve the quality of green innovation and increase stakeholders' confidence in the enterprises' green development (Xu et al. 2016).
On the one hand, environmental target constraints increase the rigidity of environmental regulations. Enterprises must comprehensively and transparently disclose environmental information, which will guide investors to pay attention to enterprises' green technical innovation and applications and help enterprises acquire financial support for their green technical R&D through legitimate means ). On the other hand, under the environmental target constraint system, green credits, carbon financing, and other market-oriented financial instruments keep improving and growing, which under influence of relevant policies, were connected extensively with enterprises' green innovations (He et al. 2021). Additionally, the environmental accountability system boosts public confidence in enterprises' environmental expenditures and signals investors to provide support to green innovations, which indirectly helps mitigate the financing constraints enterprises face in carrying out green innovations (Pan et al. 2021a, b). After mitigating their financing constraints, enterprises under pressure from mandatory environmental regulatory tools like compulsory pollutantdischarge standards and fees will be forced to carry out technical innovations to control their pollutant-discharge costs and save environmental costs. When compensation from innovations exceeds compliance costs, enterprises increase the R&D expenditures ). Simultaneously, enterprises gain a competitive advantage and reduce the energy consumption and emissions through technological innovation, which is more helpful for enterprises to benefit under a market-oriented environmental regulation system, such as the emission trading mechanism , which further motivates enterprises to increase their R&D expenditures. Increased R&D expenditure and innovation input will ensure enterprises' continuous and steady participation in high-quality green innovations with longer cycles and deeper and broader influence.
Therefore, we propose: Hypothesis 4: Local environmental target constraint intensity will press enterprises to improve green innovation quality by increasing their media exposure, enhancing their awareness of social responsibility, boosting their R&D expenditure intensity, and mitigating their financing constraints.
Under pressure from environmental accountability and the integration of the digital economy with the real economy, digital transformation becomes a win-win choice for enterprises to address both profitability and environmental protection (Yuan et al. 2021). Enterprises' digital transformations will optimize the energy consumption structure , productive techniques and processes, production efficiency, and R&D efficiency (Lin and Zhou 2021) and facilitate knowledge spillover ), which will effectively drive enterprises to extend the breadth of their knowledge of green innovations. However, as local environmental target constraint intensity increases, local governments are inclined to crowd out S&T expenditure to enhance environmental performance more directly and quickly to increase environmental governance expenditure, which is adverse to the digital transformation of enterprises (Wu et al. 2021a, b). Moreover, the process of digital transformation will slow down when transformation becomes increasingly difficult, and overdue renewal of infrastructure and technologies will leave enterprises at a lower level of digitalization. The positive effect of local environmental target constraint intensity on enterprises' digital transformation will marginally and progressively decrease, which gradually slows enterprises' green innovation quality improvement. Therefore, we propose: Hypothesis 5: There is an inverted U-shaped relationship between local environmental target constraint intensity and the degree of enterprise digitalization, which creates the effect of local environmental target constraints on the quality of enterprise green innovation, which also follows an inverted U-shaped curve.

Empirical design
To investigate the influence of local environmental target constraint intensity on the quality of green innovation of enterprises, we match the data of listed enterprises according to the prefecture-level city in which they are registered to the prefecture-level city data. We construct the following fixed effect model: where GQ it represents the quality of green innovation; SETC it represents local environmental target constraint intensity; CV it represents the control variables; i and t indicate the company and year, respectively; f represents the enterprise's industry; j represents the enterprise's prefecturelevel city; d f represents industry fixed effects; d j represents city fixed effects; d t represents time fixed effects; and it represents the random error term.
According to the theoretical analysis above, we can conclude that the impact of local environmental target constraint intensity on enterprise green innovation quality may have an inverted U-shaped change trend. Hence, we add the quadratic term (SETC it 2 ) of constraint intensity of local environmental objectives into the benchmark regression model, as follows:

Variables
Enterprises' green innovation quality Currently, a large number of scholars measure the quality of patents by the number of citations. However, the number of citations is influenced by the length of time the patent has been in existence and the popularity of the applicant in the industry. Akcigit et al. (2016) provide us with such a new perspective of measurement -the knowledge breadth method. The enterprise patent knowledge breadth mainly refers to the complexity of the knowledge contained within a certain patent. On the one hand, as an essential carrier of innovative knowledge with significant economic value, the complexity of the knowledge contained in a patent will certainly affect the quality of the patent; on the other hand, the more complex the knowledge contained in a patent, the more complicated it is to imitate and improve the patented product, which will certainly affect the monopoly power of the innovative product obtained by the enterprise relying on the patent protection system and in turn profoundly affect the enterprise performance. Therefore, the breadth of knowledge for patents measured by (1) the main patent classification number is more objective than the number of patent citations to reflect patent quality. Following , we measure knowledge width by the main classification numbers of listed companies' green patents to reflect their green innovation quality. China's International Patent Classification (IPC) follows the format "section-class-subclass-main group-subgroup," for example, A01B01/10. Specifically, the first letter of an IPC denotes one of the eight IPC sections, the 2nd and 3rd digits denote the IPC class, and the 4th letter denotes the IPC subclass. As in the Herfindahl-Hirschman Index, we can define enterprises' patent knowledge width at the main group level as follows: where Z imt is the cumulative number of green invention patents and utility model patents applied for by enterprise i under the main group m up to year t, and Z it is the cumulative number of green patents applied for by enterprise i under all main groups up to year t. The higher the value of GPK it , the greater the knowledge width of enterprises' green patents, and the better their green innovation quality. For the regression analysis, we add 1 to GPK it and take its logarithm. We should note that we choose only green invention patents and utility model patents because appearance design patents adopt different numbering methods from the invention and utility model patents and thus require a different calculation method, and compared to invention and utility model patents, appearance design patents do not reflect the complexity of knowledge the enterprise used during the patent-creation process and cannot adequately embody enterprises' self-innovating capability.
For the robustness check, we refer to Li and Zhao (2020) and measure green innovation quality by the number of citations of green patents. Specifically, we use the number of citations two years after the application (GPC it ) to measure enterprises' green innovation quality. We add 1 to (GPC it ) and take its logarithm (lnGPC it ) to perform the robustness check. As there are limited years of data, we calculate the number of citations 2 years after the application only for 2011-2018.

Local environmental target constraint intensity
The central government sets the initial environmental targets and then delegates the environmental targets to provincial governments, who in turn delegate them to municipal governments for execution. The lower governments report to the higher government, and the higher governments have the right to hold the lower governments accountable. Based on the vertical distribution of environmental targets and accountability in environmental targets assessment, first, as in Yu et al. (2020), we collected and sorted the government work reports of every city for each year and manually retrieved and arranged the specific environmental targets mentioned in the government work reports and fulfillment of the targets mentioned in various public reports to gauge local environmental target constraint intensity. Following , we divide local environmental target constraints into direct environmental target constraints and indirect environmental target constraints. The former refers to the targeted emission reduction of sulfur dioxide, oxynitrides, chemical oxygen demand, ammonia nitrogen pollutants, and so on, specified by the local governments in their government work reports. 1 We use specific indicators of PM 2.5 density, PM 10 density, days of good air quality, environmental quality, and so on, though unspecified indicators of sulfur dioxide, oxynitrides, and ammonia nitrogen pollutants, among others, are still indirect environmental target constraints. 2 In contrast to environmental target constraints, environmental target constraint intensity directly reflects governments' attentiveness, input, and executive power in meeting environmental targets. The calculation of the target constraint intensity takes into account both the target setting and the actual achievement of the target. Based on the constraint intensity analysis of Liu and Huang (2019), we use the ratio of the cities' actual fulfillment of environmental targets to the original environmental targets as an indicator to measure local environmental target constraint intensity, 3 which includes both direct environmental target constraint intensity (SDETC it ) and indirect environmental target constraint intensity (SIETC it ). The actual achievement of environmental targets can directly demonstrate the implementation strength of the government's environmental governance behaviors and policies, which is why the ratio of the achievement of environmental targets to environmental targets can objectively reflect the intensity of environmental target constraints.
Only when local governments specified concrete environmental targets in the year (including specific pollutants and their targeted values) in their government work report and reported their fulfillment of the targets (including specific pollutants and actual fulfilled values) in the next year's government work reports can we calculate 1 For example, the government work report of Yangquan City in 2012 has specific emission-cutting targets of "to reduce emission of COD, ammonia nitrogen, SO2, soot and dust by 1.3%, 1%, 2%, 3%, and 3%," respectively. Its government work report of 2013 released the fulfillment of the emission-cutting targets: "emission of COD, ammonia nitrogen, SO2, soot and dust has been reduced by 4.18%, 0.78%, 7.12%, 5.96%, and 5.1%," respectively. This example is typical of direct environmental target constraints. 2 For instance, the government work report of Benxi City in 2017 set the environmental target of "80% of days in a year will have good air quality", and its government work report of 2018 (released in 2017) reports 318 days of good air quality and the environmental target is fulfilled. Therefore, we see an indirect environmental target constraint. 3 In calculation, we need to unify the units used in specific targets and their fulfillment values. For example, we need to unify the tons of emission reduction and percentage of emission reduction and unify the number of days with good air quality and the percentage of days with good air quality and so on. Additionally, if the government work report mentions that its specific environmental target is to fulfill the same provincial environmental target, then given the environmental accountability between the provincial and municipal governments, we set the environmental target constraint intensity to 1.
the corresponding constraint intensity. If local governments did not release specific environmental targets, but reported fulfillment in the next year's work report, we then set the year's environmental target constraint intensity to 1; if local governments neither set concrete environmental targets nor released fulfillment statements or they disclosed only environmental targets but did not report fulfillment, then we set their environmental target constraint intensity to 0. We have conducted preliminary statistics on the environmental target constraints in the government work report of each prefecture-level city from 2011 to 2019, as shown in Table 1. We matched the data of A-share-listed enterprises on the Shanghai Stock Exchange and Shenzhen Stock Exchange from 2011 to 2019 to their green patent applications. From these samples, we removed enterprises without green patent applications from 2011 to 2019, finance and insurance enterprises, ST and PT enterprises, and insolvent enterprises. We excluded the sample of companies with serious missing data and those with the number of years of green patent application not more than 3 years in the sample period and used the linear interpolation method and weighted moving average method to fill in some of the missing data to obtain the data of 1836 listed companies, with a total of 8851 research samples. We obtain the macro-data of the cities in which they are registered. Listed enterprises' green patent classification data are from the State Intellectual Property Rights Bureau. We retrieved information on green patent citations from the database of listed enterprises' operational data from the China National Research Data Service Platform (CNRDS). We collected and sorted the data to calculate environmental target constraint intensity and environmental regulation rigidity manually from the cities' original government work reports during the sample period. Enterprises' digitalization level data are from their yearly reports. The social responsibility data of enterprises are from Hexun Net. We obtained enterprises' media attention data from the Financial News Database of Chinese Listed Companies. The data for the other variables and calculations are from the China Stock Market & Accounting Research (CSMAR) database and the WIND database. Data for the remaining city-level variables are from the China City Statistical Yearbooks. We matched the local environmental target constraint intensity and citylevel variables with enterprises' panel data according to the cities in which the listed enterprises are registered. Except for financial constraints and the nature of shareholding, we take the logarithms of the variables. Table 2 lists the core variable definitions and calculation methods. Due to space

Panel unit root tests
Before regression analysis of panel data, a unit root test is needed to ensure data stability and avoid false regression problems. Since the panel data used in this paper is an unbalanced panel, two methods, IPS test and Fisher-ADF test, are used in this paper for panel unit root test. The results of the panel unit root test are shown in Table 3. It can be seen that no matter whether it is the IPS test or Fisher-ADF test, all variables reject the null hypothesis that there is a unit root, that is, all variables are stationary.

Influence of local environmental target constraint intensity on enterprises' green innovation quality
According to Table 4, the coefficient of the influence of direct environmental target constraint intensity on lnGPK it is not prominent, which indicates that the incentive effect of direct environmental target constraint intensity on enterprises' green innovation quality is not strong. The reason may be that direct environmental target setting focusing on specific pollutant emission reduction will prompt local governments to prefer command-based environmental regulation tools to meet mandatory emission reduction targets, and because of the existence of penalty caps and limited  innovation compensation from improving the quality of green innovation, some enterprises are satisfied with "meeting the standards" of environmental standards. Moreover, when enterprises cannot obtain more revenue through green innovation, they will abandon their investment in high-quality green innovation and turn to strategic and responsive green innovation or even low-quality green innovation with lower cost and difficulty to gain short-term benefits. In addition, based on the significance of the regression results of the quadratic term coefficients, it is clear that there is no significant nonlinear characteristic in the effect of the intensity of direct environmental target constraints on the quality of corporate green innovation. This differs from the findings of Yu et al. (2020). The reason is that Yu et al.'s (2020) study examines the quantity of patents, while this paper focuses on the quality of green patents of enterprises. This also indicates that there are some discrepancies in the effects of local environmental goal constraints on patent quantity and patent quality. The coefficient of indirect environmental target constraint intensity on lnGPK it is significantly positive, which shows that setting indirect environmental targets gives enterprises more freedom to choose and stimulates active green innovation. This differs from the findings of Tao et al. (2021). The reason for this is that the innovation capabilities of firms in the sample used by Tao et al. (2021) vary widely, and the enterprises with weak innovation capabilities are the main group causing the decline in the quality of innovation activities, while firms in the sample used in this paper are Chinese-listed companies with relatively strong innovation capability, which in turn causes the difference in the research results. However, the quadratic term coefficient is − 0.055 at the 1% significance level, indicating that indirect environmental target constraint intensity has an inverted U-shaped relationship with enterprises' green innovation quality. Initially, indirect environmental target constraint intensity will stimulate high-quality green innovations, but at a certain level of green innovation quality, enterprises faced with higher indirect environmental target constraint intensity will have decreasing marginal income from new, higher R&D investment as green innovation become progressively harder. Therefore, the motivation for enterprises to improve green innovation quality weakens continually. The positive influence of indirect environmental target constraint intensity on enterprises' green innovation quality decreases marginally and progressively, which is in line with Hypothesis 2. Specifically, when lnSIETC it reaches 1.182, the inverted U-shaped curve reaches its apex. According to the variable statistics, when indirect environmental target constraint intensity exceeds the critical value for only very few samples, indirect environmental target constraint intensity still has stimulating effects on enterprises' green innovation quality.

Endogeneity and robustness tests
We conducted robustness tests from three aspects: replacing the explained variables, accounting for city heterogeneity, and changing the regression methods. Specifically, (i) we replace lnGPK it with lnGPC it ; (ii) we remove sample enterprises from cities with better environmental quality, like Qingdao, Yantai, Lishui, Taizhou, Fuzhou, Xiamen, Shenzhen, Zhuhai, Huizhou, Zhongshan, Guiyang, and Kunming and rerun the regression; and (iii) using a panel interactive fixed effect model we control for fixed effects such as industry, city, and year and further control the interactive industry-city, industry-year, and city-year fixed effects. Additionally, according to the requirement of correlation and exogeneity of instrumental variables, we (iv) use river density and annual precipitation as instrumental variables for the intensity of environmental target constraints and its quadratic terms, and conduct two-stage least squares (2SLS) regressions on the intensity of local environmental target constraints and the quality of corporate green innovation. For the first three tests, we report the robustness test results in Table C.2. We provide the results of the instrumental variable estimation in Table C.3. The test statistics confirm the existence of endogeneity and the validity of the instrumental variables. The direction and significance of the regression coefficients from the robustness test and IV-2SLS regression are generally consistent with the baseline regression results, indicating strong support for the original findings.

Mechanism tests
The previous analysis concludes that direct environmental target constraint intensity has an insignificant influence on enterprises' green innovation quality; therefore, we analyze the transmission mechanism from only indirect environmental target constraint intensity to enterprises' green innovation quality. Based on models (1) and (2), we build mediation models. Models (3) and (5) describe the influence of indirect environmental target constraint intensity on the intermediate variables, while models (4), (6), and (7) depict how indirect environmental target constraint intensity influences enterprises' green innovation quality through the intermediate variables.

City-level tests
Based on the previous analysis of the theoretical mechanism, we test the two mechanisms of environmental regulation rigidity and governmental S&T expenditure. As Table C.4 shows, indirect environmental target constraint intensity prominently enhances the rigidity of environmental regulations and forms an inverted U-shaped relationship with enterprises' green innovation quality through environmental regulation rigidity, which is similar to the findings of Pan et al. (2021a, b). When environmental performance is included in the performance assessment, local governments pay more attention to environmental protection, i.e., environmental target constraint drives up the intensity of environmental regulations. The increase in the intensity of environmental regulations will make the marginal cost of green innovation rise, leading to a declining innovation compensation effect and a lower innovation incentive effect, which in turn will have a marginal decreasing positive impact on (3) ln GQ it = 0 + 1 (ln SIETC it ) 2 + 2 ln SIETC it + 3 (ln IMV it ) 2 + 4 ln IMV it the quality of green innovation of enterprises. Moreover, indirect environmental target constraint intensity deterred governments from making more S&T expenditures due to the requirement for more environmental spending, which negatively affected enterprises' green innovation quality. This is in line with Hypotheses 2 and 3.

Enterprise-level tests
We test the five micro-mechanisms of enterprises' media attention, social responsibility, R&D expenditure intensity, digitalization level, and financing constraints. According to Tables C.5. and C.6., indirect environmental target constraint intensity positively affects enterprises' green innovation quality by increasing negative news reports, strengthening their social responsibility, and boosting their R&D expenditure intensity. The results of the test on R&D expenditure intensity are similar to the findings of Yu et al. (2020). Then, indirect environmental target constraint intensity will relieve enterprises of financing constraints to encourage them to improve green innovation quality. This is because the environmental target constraint increases the concern of society as a whole about enterprises' emission behavior and generates a signaling effect, which indirectly reduces the difficulty of financing green innovation by enterprises.
Enterprises will actively fulfill their social responsibility for better development and increase R&D investment to reduce pollution emission under the policy guidance and social opinion supervision. On the other hand, although the digital transformation of enterprises is one of the effective ways to cope with the government's environmental target constraints, the decline in governmental S&T expenditure and the high complexity, expense, and uncertainty of digital transformation itself will hinder the process of digital transformation of enterprises, which will lead to an inverted U-shaped relationship between indirect environmental target constraint intensity and enterprises' digitalization level. Because the digitalization level has positive effects on green innovation quality, indirect environmental target constraint intensity, by influencing the digitalization level of enterprises, has a steeper inverted U-shaped relationship with their green innovation quality. The test results comply with Hypotheses 4 and 5.

Change in enterprises' expected net income
To determine the source of the inverted U-shaped relationship between indirect environmental target constraint intensity and enterprises' green innovation quality, based on the theoretical mechanism and model analyses above, we continue using the intermediate effects model to observe changes brought about by boosted R&D input on expected net income under increasing indirect environmental target constraint intensity. Sales profit is the major source of enterprise income; therefore, it can accurately reflect their income level. We add 1 to enterprises' sales revenue growth rate (RGR it ) and net profit margin (NPM it ) and take their logarithms (lnRGR it , lnNPM it ). These explained variables represent enterprises' expected net income level. We use lnR&D1 it and lnR&D2 it as intermediate variables. Table 5 summarizes the regression results. The influence of lnSIETC it on lnRGR it is significantly negative, as is the influence of lnR&D1 it and lnR&D2 it on lnRGR it . Considering the results of the mechanism analyses, we find that enterprises facing increased indirect environmental target constraint intensity will increase their R&D input, which will reduce the growth rate of sales revenue. This is mainly because as the intensity of the indirect environmental target constraint drives the increase in R&D investment, leading to an increase in the production and operation costs, coupled with the limited direct contribution of green innovation to sales revenue, the growth of sales revenue from green innovation has a marginal decreasing trend. Next, the coefficients of lnR&D1 it and lnR&D2 it in relation to lnNPM it are significantly negative, which further proves that raised indirect environmental target constraint intensity will encourage enterprises to increase R&D expenditure intensity and consequently bring about a marginal and progressive decrease in their expected net income. After further observing the nonlinear influence of R&D expenditure intensity on the net sales profit margin, we find that the regression coefficients of (lnR&D1 it ) 2 and (lnR&D2 it ) 2 in relation to lnNPM it are significantly positive, which proves the existence of a prominent U-shaped relationship between R&D expenditure intensity and net sales profit margin. Due to the long transition period of green patents, the increasing complexity of green innovation, and the cost of R&D investment, especially high-quality green innovation, the return on sales revenue from R&D investment is not obvious in the early stage of R&D. When the intensity of R&D expenditure and technology level reaches a certain level, the technology and achievements from the increased intensity of R&D expenditure mature and are converted into applications, which will significantly promote the increase in the net sales margin. In summary, the main reason behind the marginal and progressive decrease in the positive influence of indirect environmental target constraint intensity on enterprises' green innovation quality is that when enterprises face pressure from indirect environmental target constraint intensity, they will raise their R&D expenditure intensity to improve green innovation quality, which will consequently reduce their expected net income and weaken the incentive to improve green innovation quality.

Regional heterogeneity analysis
We first examine age heterogeneity when local officials face promotion incentives. According to the previous analysis, it is known that the performance of local officials' environmental target constraint assessment will directly affect the promotion of officials, and the addition of strong environmental target constraint to the original economic target assessment system has transformed the behavior of local governments. Therefore, the different promotion incentives and promotion pressures faced by local officials will lead to discrepancies in the environmental pressure exerted by local governments on enterprises, which in turn will have a heterogeneous impact on the quality of their green innovation. Considering that the probability of promotion gradually decreases as the age of local officials increases and the chance of promotion for local officials over 57 years old decreases significantly, this paper refers to Yu et al. (2020) and divides the sample into two groups of mayors less than or equal to 57 years old and older than 57 years old for heterogeneity analysis, using whether the mayor is over 57 years old as the threshold. According to Table 6, mayors under 57 years of age will be more easily stimulated to obtain a promotion. Local governments and their officials are more proactive in implementing policies and measures; therefore, indirect environmental target constraint intensity has significantly positive effects on enterprises' green innovation quality. In contrast, in areas with mayors aged above 57 years, because of the lack of adequate promotion incentives and pressure to advance, the relevant departments lack patient encouragement, supervision, and review of patent quality, and environmental target constraint policies may not be implemented thoroughly. Therefore, their stimulating effect on enterprises' green innovation quality is insignificant. This is similar to the findings of Yu et al. (2020). According to timeline of environmental target assessments written into the municipal and provincial government work reports, we classify environmental target constraint policies into policies "prioritizing city over province" and policies "prioritizing province over city." In the former case, local officials are more proactive and internally motivated to roll out and firmly implement relevant policies and the qualitydriven green development strategy. However, in the latter case, local officials in most cases will passively accept mandatory requests from upper authorities. Therefore, policies are better implemented in areas with policies "prioritizing city over province" than in areas with policies "prioritizing province over city."

Industry heterogeneity analysis
The industries of the sample enterprises are classified into technology-intensive, capital-intensive, and labor-intensive industries. Table 7 reports the regression results for three different types of industries, which show that indirect environmental target constraint intensity has a significant positive effect on the green innovation quality of enterprises in technology-intensive industries and a significant inverted "U" curve relationship with the green innovation quality of enterprises in capital-intensive industries, while it has little effect on labor-intensive industries. Technology-intensive and capital-intensive enterprises with a certain technology and capital base have relatively low cost and challenge to carry out green innovation under the policy and institutional guidance of environmental constraints, while high-quality green innovation is beneficial for these enterprises to obtain higher compensation for innovation. The sample industries are then divided into heavily polluted industries and moderately lightly polluted industries. According to Table 8, there is a significant inverted "U" curve relationship between indirect environmental target constraint intensity and green innovation quality of enterprises in medium and light pollution industries, which is unable to drive the improvement of green innovation quality of enterprises in heavy pollution industries. The reason is that the overall level of green innovation of enterprises in heavily polluting industries is relatively poor, the complexity and cost of enterprises to obtain innovation compensation effect through green innovation is excessive, as well as the limited resources and foundation for green innovation, which in turn makes it hard to sustain high-quality green innovation activities.

Conclusions and policy suggestions
Based on the unique characteristics of China's environmental responsibility system, this study innovatively constructs local environmental goal constraint intensity indicators based on actual evidence of local government environmental target constraint assessments by combining environmental policies with government environmental responsibility assessments. Based on a theoretical model that explores the impact of environmental target constraint intensity on enterprises' green innovation behavior, we empirically test the impact mechanism of different types of local environmental target constraint intensity on the quality of enterprises' green innovation. The analysis yielded several findings.
Indirect environmental target constraint intensity has significantly positive effects on enterprises' green innovation quality, though they have an inverted U-shaped relationship after further observing their nonlinear characteristics. The positive effects of indirect environmental target constraint intensity on enterprises' green innovation quality decrease marginally and progressively; however, currently, the impact of indirect environmental target constraint intensity on enterprises' green innovation quality is mostly facilitated. Direct environmental target constraint intensity that aims to reduce the discharge of specific pollutants has insignificant effects in improving enterprises' green innovation quality.
In the mechanism tests, indirect environmental target constraint intensity has an inverted U-shaped trend by increasing the strength of environmental regulation and influencing the degree of digitalization of enterprises, which in turn forms an inverted U-shaped relationship with the quality of green innovation. In addition, the intensity of indirect environmental target constraints will not only promote the quality of corporate green innovation by increasing corporate media attention, corporate social responsibility, and R&D expenditure intensity and alleviating financing constraints, but will also negatively affect the quality of corporate green innovation by suppressing government S&T expenditure. Further analysis reveals that, in the face of the indirect environmental target constraint intensity, the increase in R&D expenditure intensity for green innovation quality improvement decreases the expected net income, resulting in insufficient incentive for innovation quality improvement, which is an essential reason for the marginal decreasing trend of the positive impact of the indirect environmental target constraint intensity on the quality of green innovation of enterprises. Simultaneously, environmental target constraints work better in areas with policies prioritizing cities over provinces and with mayors under 57 years of age than in areas with policies prioritizing province over city and with mayors over 57 years old. The indirect environmental target constraint intensity has a significant positive impact on the green innovation quality of technology-intensive industries, with an inverted "U"-shaped relationship with the green innovation quality of capital-intensive industries and medium and light pollution industries.
Based on these conclusions, we can propose the following suggestions to encourage enterprises to improve green innovation quality, explore new paths of high-quality economic development with a high level of ecological environment protection, and seek new dynamics of economic growth: Improve the design of policies and systems Deepen the reform of the environmental target accountability system, perfect the environmental target constraint system, and take steps to establish and improve the environmental target constraint system based on indirect environmental target constraints and complemented by direct environmental target constraints. Attach special importance to assessing the fulfillment of indirect environmental targets and release the fulfillment in the next year's government work report. Meanwhile, there should be more effort to review enterprises' patents, scientifically and prudently patent authorization management systems to prevent enterprises from conducting tactically contingent green innovations.

Inspire the internal motivation of enterprises
We should diversify the mono-system that is based on mandatory environmental regulations, strengthen the market orientation of green innovations, and improve market-oriented environmental regulation instruments such as emission rights trading mechanisms to increase compensation for enterprises' green innovations and mitigate the rising marginal cost of green innovation for firms due to the increased intensity of environmental regulations, which could increase the incentives for improving green innovation quality to improve the inverse "U"-shaped relationship between the intensity of environmental regulations and enterprise's green innovation quality. Meanwhile, local governments should be guided to formulate strategies for long-term development by S&T innovations and increase fiscal S&T expenditures. In addition, the media attention on enterprises should increase appropriately to enhance corporate social responsibility, and the investment and policy support for the digital transformation of enterprises should be increased to enhance the technical foundation and endogenous power of enterprises' green innovation quality improvement.
Optimize policy implementation effects Local governments should expand their support and subsidies for corporate R&D activities, technological transformation, and upgrading, especially for challenging and high-quality green innovation research projects of enterprises in heavily polluting industries, so as to alleviate the decline in expected net income resulting from increased R&D expenditures by companies in the face of environmental target constraints. To deepen the reforms to streamline administration, decentralize power, and improve services, local governments should consider local environmental quality and green technology development to flexibly and proactively roll out a complete set of detailed environmental targets adapted to local conditions, shoulder the corresponding environmental responsibilities. Meanwhile, localities can cultivate more young and knowledgeable cadres to be willing to breach the inefficient tradition of promoting based on seniority and other regular promotion paths, and breaking the constraints on departments, sectors, systems, and identifications so young cadres can make the most of their talents.
Although this paper comprehensively analyzes the impact of local environmental target constraints on the quality of enterprise's green innovation and the nonlinear characteristics of its impact, there are still some limitations, which could also be possible future research directions. For example, due to the limitations of micro-firm data, it is challenging to conduct a comprehensive examination of the threshold effect between local environmental target constraints and enterprise's green innovation quality, which is an extremely theoretical and practical research project in the future. Additionally, due to the strategic interaction behavior of environmental regulations among local governments and the intraregional production factor flows caused by the variations in the intensity of environmental regulation enforcement in various regions, it would make the impact of environmental target constraints on the quality of enterprise's green innovation may have spatial spillover effects. We did not explore the spatial effect between environmental target constraints